FW: MacArthur-ref. of low validity/reliability?

[Dale, Philip S.]

Several people asked me if I had responded to the enquiry from Amy Khasky (included below). I had thought my response went to the entire list, but apparently not, so I am re-sending it (see below). Ms. Khasky and I have corresponded directly since then; her research involved comparing performance on some experimental tasks with the suffixes portion of the CDI:WS (irregulars and overregularizations). Almost all the information on reliability and validity of the CDI:WS has been focussed on Vocabulary Production and Sentence Complexity. The portions of the instrument looking at suffixes were included for research purposes. There are several important publications of Bates, Marchman, and others on those measures. Those publications, and the coherent pattern of results they present, provide a kind of convergent validation. But there really aren't any direct comparisons of those sections with comparable measures based on language samples, to my knowledge. 
For the irregulars, I wouldn't expect much of a correlation with a learning task, since the whole point of irregulars is that they are learned as separate lexical forms, hence there shouldn't be much generalization from the ones you already know (that's a slight oversimplification, to be sure, given the existence of partially regular irregulars, but I think the main conclusion holds).
Overregularizations are different, since they are all about patterns, of course. But it's important to keep in mind that although the developmental trend is increasing over the norming age range of the CDI:WS (16-30 months), ultimately this is a curvilinear trend - children diminish and eventually stop producing overregularizations. If the measure is used up to 36 months, a curvilinear trend may be beginning, while the pearson correlation coefficient looks only for linear trends. 

Philip Dale
==============================
This is a somewhat oddly-worded enquiry. Certainly the MacArthur (now the MacArthur-Bates CDI) works better in some situations than others. For example, the vocabulary comprehension measure can be problematic at early ages (up to a year or so), especially for lower-SES parents. Concurrent validity correlations are highly variable, depending on the criterion measure used, which probably says more about the complexity of language than about validity per se. And predictive correlations are generally lower than concurrent ones, though that says more about the genuine variability of growth trajectory than anything else. 
As a one-word and generally accurate response to the question, I do like "solid."  But as it happens, I have been collecting all published studies that have used the CDI in order to prepare the section on reliability & validity for the forthcoming revised edition of the CDI manual. I invite Ms. Khasky to correspond with me directly to see if I can be of help.

Philip Dale
Communication Science & Disorders
University of Missouri-Columbia
dalep at health.missouri.edu

-----Original Message-----
From: info-childes at mail.talkbank.org on behalf of amy khasky
Sent: Sun 10/31/2004 02:04
To: info-childes at mail.talkbank.org
Subject: MacArthur-ref. of low validity/reliability?
 
Hello,

I'm having trouble finding any references that have found low reliability 
and/or validity ratings for the MacArthur.  Is it just a solid measure?

Thanks in advance for any help,
amy khasky